Generic object recognition using 2D PCA and virtual manifolds
P. Deepti, Saurabh Das · 2006
This paper presents a method for generic object recognition invariant to pose and a wide range of illuminations using virtual objects and their manifolds. To account for these variations, the algorithm simulates the process of rendering a virtual model, using computer graphics. Appearance information from the synthesized images of the virtual model are used to learn the variations in the multiple view feature descriptors using 2D PCA. Recognition problem is approached by using a hierarchical framework, which involves different stages of processing. First an intelligent generic recognizer based on 2D PCA is employed to reduce the search space to a few rank ordered samples. Average of eigen distances and distance transform based correlation is then used to verify the correct object class. We have achieved very good recognition rates on the IITM generic solid object library (IGSOL) which we created.